HaiAI123

Curated Global AI Tools Directory

Industry update
Flat

Ex-DeepSeek Researcher Zhang Bo Open-Sources Local Agent Linger with 2k GitHub Stars

11/20/2026 — 11/21, 18:18·3 sources (all)·3 reports (all)

Key takeaways

  • Zhang Bo, former DeepSeek-VL lead and founder of Simplex AI, has open-sourced Linger, a personal Agent built to run locally on a user's own…

Background & analysis

On November 19, 2026, Zhang Bo, former lead of the DeepSeek-VL project, announced on X the open-source release of his personal Agent project, Linger, with the code uploaded to GitHub. The project is developed by Simplex AI, a company Zhang founded. According to the project documentation, Linger can read local email, calendar entries, and files, and can operate a browser the user is already logged into. Read actions run directly, while write actions such as sending email trigger a confirmation card that the user can take over at any time. Linger restricts Agent actions through confirmation cards and server-side verification, and does not use sandbox isolation. Simplex AI states that local inference produces no cloud billing and no token costs, and reports a generation speed of 45.5 tokens/s for the Alibaba Qwen3.8-Flash-Next model on a 12GB VRAM and 64GB RAM setup, noting that this figure is the steady-state speed after generation starts and excludes first-token latency. That performance claim has been flagged as coming from differing testing conditions.

On November 20, 2026, further coverage identified Zhang Bo, also known as Tony Zhang, as a former DeepSeek researcher and Simplex AI founder. In his post, Zhang compared the release to the arrival of the personal computer, arguing that Agents should stop depending on cloud data centers and run directly on users' own machines, letting users own their data and personal context. The project drew quick attention, with GitHub stars surpassing 2k at the open-source address github.com/simplexai-labs/linger. The reports also introduced Linger's proactive assistant, Another Me, which continuously monitors a user's inbox, calendar, and Slack, unlike most Agents that wait for explicit user instructions. On November 21, 2026, one more report described the work as a startup project by an author of DeepSeek V1, focused on local deployment, without disclosing model scale, benchmark results, or the exact scope of the release. The matter currently stands at the open-source announcement stage, with performance claims still open to verification.

AI-generated from 3 reports · updated 15 hours ago

Latest turnQuantumBit reports that the author of DeepSeek V1 has revealed a new startup project: an open-source agent designed for local deployment. Details on model size, benchmark results, and release scope were not disclosed.

Related AI tools

updream
24-hour heatHeat index 114 · peak 201 · 19h ago
24 hours agonow

Reports on this story headlines open the original

Yesterday
  1. QuantumBit reports that the author of DeepSeek V1 has revealed a new startup project: an open-source agent designed for local deployment. Details on model size, benchmark results, and release scope were not disclosed.

    量子位AI score 52
Nov 20
  1. Tony Zhang, founder of Simplex AI and a former DeepSeek researcher, has open-sourced Linger, a personal agent designed to run on a user's own computer instead of cloud data centers; the GitHub repo has passed 2k stars, Zhidx reported. Its proactive assistant, Another Me, continuously watches the user's inbox, calendar and Slack.

    智东西AI score 57
  2. Zhang Bo, former lead of the DeepSeek-VL project, announced on X that he has open-sourced Linger, a personal agent built by his company Simplex AI, with the code now on GitHub. Linger reads local email, calendars and files and drives the user's logged-in browser; read actions run directly while writes such as sending mail require a confirmation card, and it relies on those cards plus server-side checks instead of sandboxing. Simplex AI says local inference incurs no token fees, though its 45.5 tokens/s figure on a 12GB-VRAM setup comes from lab tests with differing baselines.

    华尔街见闻AI score 56

Other stories people are talking about

How is heat calculated?About the method

Heat counts how many independent sources covered a story in the last 48 hours: one source counts once no matter how many posts it published, decaying with a 24-hour half-life. What ranks first is what many people are talking about.

This page aggregates public feeds. Headlines and summaries are machine-organized and remain the property of the original authors; verify important facts at the source. If you believe a headline or summary infringes your rights, email the contact address in the footer with the page URL and basis of your claim, and we will remove or replace it after review.

Surge
Discussion rising fast
New
First report within 6 hours
Rising
Still gathering discussion

Back to AI industry updates →